From: John Conover <john@email.johncon.com>
Subject: Re: The Standard: Warning: Recession Ahead
Date: 7 May 2001 00:17:25 -0000
And what about the dynamics of the dot-com industry, as mentioned below:
Its not so much that static methodologies, (like MBA/MBO,) don't
work in organizing high entropy dynamic systems, (which the
dot-coms are/were,) as it is that they lead to irrational,
unattainable, expectations.
and:
As approximate numbers, (taken from the average of many industries
on http://www.johncon.com/ndustrix/,) to exploit the lack of
predictability in the US GDP and its constituent industrial
markets, the short term predictability of a few months to a 70%
accuracy means that about (2 * 0.7) - 1 = 40% of a company's
assets should be at risk in the short term, in WIP, capital
expenditures, R&D, etc.
If the short term management horizon is a month, (which it has to be
to get a maximally optimal 2X growth per year,) then, on average, a
management paradigm, (prediction, and execution,) would be obsolete
where 1 / sqrt (t) = 0.5, or about four months-i.e., half would be
obsolete in less than four months, half would last longer.
About 30% of the paradigms would have to be changed by the end of the
second month, 58% by the end of the third month, and half by the end
of the fourth month.
That's how fast the executive suite has to move the company-which is a
continual struggle in management trails and tribulations. Its what it
takes to sustain a 2X annual corporate growth. Quarterly planning
isn't a viable option, (which is the typical MBA/MBO planning
cycle-which, although workable in old economy smokestack/assembly line
industries, isn't applicable to contemporary service industries.)
To achieve 2X growth requires a managed, (as opposed to an
administered,) solution.
And that's where the dot-coms went wrong.
John
BTW, the distributions, (they are from fractal dynamics,) have very
"fat tails". (They have nothing to do whatsoever with Gaussian/Normal
distributions, and are a completely different family.) All is not
always over by the end of four months, however. What's the chances
that a management paradigm will last a century of months?
Its about 3%, (1 / sqrt (1200) = 2.9%.) Those are the companies in
industries that become the mainstay of the GDP. About 3% of the
industries in the last century, (including oil, rail, automotive,
electronics, etc.,) achieved that status. The natural evolution of
companies and industries is a log-normal distribution; where the 3%,
or so, dominate the GDP, (like 70% of it, or so.) They end up
generating almost all the wealth.
John Conover writes:
>
> As another interesting sidebar on project management, many of the
> projects attempted in the dot-coms were executed under MBA/MBO
> methodologies.
>
> Unfortunately, about 84% failed, (that's the not-so-pretty industry
> numbers for the IT projects-failure meaning over budget, over
> schedule, or both; re: The Standish Group's, Jim Johnson,
> http://www.standishgroup.com/msie.htm, which authored
> http://www.scs.carleton.ca/~beau/PM/Standish-Report.html,
> http://www.garynorth.com/y2k/detail_.cfm/1088, and
> http://www.systemcraftllc.com/standishGrp.html.)
>
> Why are the numbers like that?
>
> If project milestone completion, c, is considered a linear function of
> resource allocation, (i.e., c = kt,) and the project's decision
> process has sufficiently high entropy, then the real milestone
> completion would be closer to c = sqrt (kt).
>
> In other words, the ratio of the projected milestone completion, to
> the real would be sqrt (kt); and for all t, the average project would
> have less than sqrt (kt) milestone's finished, for one standard
> deviation of the time, (i.e., 84.1% of the projects would be less than
> sqrt (kt) finished-where they should be kt finished-and 15.9% would be
> greater than sqrt (kt) finished.)
>
> Or, 1 / 0.841 = 1.19 standard deviations would be better than on
> schedule, or on budget, which is 11.7%; and the remaining 88.3% would
> be over budget, in one way or another, (which is less than a 5% error
> compared with the Standish metrics on the IT industry, as a whole.)
>
> Its not so much that static methodologies, (like MBA/MBO,) don't work
> in organizing high entropy dynamic systems, (which the dot-coms
> are/were,) as it is that they lead to irrational, unattainable,
> expectations.
>
> So, what would be rational expectations?
>
> One should square the resource allocation requirements obtained by
> linear estimates to get a realistic budgetary and schedule
> requirements.
>
> John
>
> BTW, you have to keep the projects small, and simple, too. Complex
> projects have resource allocation issues that explode-the square
> function starts rising very rapidly on complexity.
>
> How many complex IT projects have succeeded, (The Standish Group
> uses ten million bucks as the benchmark for a complex system?)
>
> In the history of IT, none.
>
> John Conover writes:
> > As a sidebar, VCs usually want an investment to run about 5 years,
> > then the company IPOs, and the VC makes lots of money-and so do the
> > employees, (sound like a Ponzi scheme? It is.)
> >
> > What's the VC industry's success rate?
> >
> > The industry numbers are about one in nine, (the remaining eight get
> > redeployed, reorganized, or re-something'ed into Chapter 7 and 11
> > which is a popular end game with the dot-coms these days.)
> >
> > What's the probability of a company's strategy and marketplace lasting
> > at least 5 years = 60 months, without having a negative month where it
> > couldn't meet the payroll?
> >
> > The deviation of the distribution of the GR of all companies at the
> > end of 5 years is about 1 / sqrt (60) = 0.13, or about 15.8% of the
> > companies will have done at least 87% of what they were supposed to
> > do. So, we need 1 / 0.87 standard deviations = 1.14, which corresponds
> > to 0.127, or about one in 8-or about a 10% error from the empirical
> > metrics published by the industry associations.
> >
> > Call it about one in ten, which is the number I used, below, in the
> > statement:
> >
> > ... We would, also, expect about a 1 in 10 survival rate, (e.g.,
> > any company that has less than 10% market share, e.g., a 90%
> > chance of failure, won't make it,) and we are seeing those
> > numbers." ...
> >
> > So, one probably shouldn't invest in companies with less than a 10%
> > market share, (or do business with them, either,) unless there is
> > compelling reason to do so, since only about 1 in 10 will survive at
> > least 5 years.
> >
> > The one in ten number is consistent with what has been published about
> > the other industrial "bubbles" of the 20'th century, (specifically,
> > both the radio and television set industry, of the 20's and 50's, the
> > electronic game industry of the mid 70's, the CB radio industry of the
> > early 80s, and the software industry of the late 80's-and now the
> > dot-com industry.)
> >
> > John
> >
> > BTW, the number is consistent with the other "bubbles" in the distant
> > historical perspective, too, including tulips in 17'th century
> > Holland, and the South Sea "bubble" of the 18'th century. As a
> > generality, about 90% of the investors lost 90% of their investment,
> > and the remaining 10% increased their wealth by about 10X, on average,
> > in each of the "bubbles" throughout the last 5 centuries. Such things
> > are Ponzi, (or pyramid,) schemes, and executive management should know
> > better than to participate, and run companies like a grand industry
> > sanctioned Ponzi scheme, (that goes for the investors, too.)
> >
> > Interestingly, according to MBA dogma, the executives in the dot-coms
> > did everything right. The MBA methodology of corporate organization is
> > not new, and it works well for what it was designed to do-efficiently
> > manage an industrial assembly line process through Taylor's
> > time-motion analysis. Extrapolating the methodology as a management
> > paradigm is a large leap of faith, however, and business schools have
> > to shoulder a large part of the blame for the dot-com "bubble" for not
> > including modern concepts into the curriculum.
> >
> > But the young executives that operated the dot-com "bubble" have a lot
> > of historical company. None other than Sir Isaac Newton lost a
> > fortune in the South Seas "bubble" of 1720:
> >
> > ... Sir Isaac Newton, scientist, master of the mint, and a
> > certifiably (sic) rational man ... sold his 7,000 pounds of stock
> > in April for a profit of 100 percent. But something induced him to
> > reenter the market at the top, and he lost 20,000 pounds,
> > ... [prompting him to say] ... "I can calculate the motions of the
> > heavenly bodies, but not the madness of people."
> >
> > Re: Harvard Magazine,
> > http://www.harvard-magazine.com/issues/mj99/damnd.html, and search
> > forward for the word "Newton".
> >
> > John Conover writes:
> > > Looking at the valuation of a well managed company, (i.e., one where
> > > management knows what its doing-knows what has to be done, and knows
> > > how to execute on it, optimally, e.g., grow 2X per year, for at least
> > > 10 years, before it starts its demise,) and if it starts with a
> > > million bucks of GR the first year, (a reasonable value,) then IPOs in
> > > year 5, (with a GR = 2^5 * 10^6 = 32 million-a reasonable number from
> > > the historical perspective,) then in year 10 it would have a GR of 2^5
> > > * 32^6 = one billion bucks.
> > >
> > > If the company was floated on an initial million bucks investment,
> > > (i.e., it was not profitable only in the first year,) it would be a
> > > thousand X increase in valuation over the decade, (assuming that the
> > > capitalization vs. GR remained constant, which was 3X in the 20'th
> > > century, averaged over all companies-I'm using a conservative 1X.)
> > > Compare that with the 100X the best dot-coms delivered, (and then
> > > didn't.)
> > >
> > > Now assume that management is less efficient. It grows the company at
> > > only 10 percent less than optimal, at 1.9X per year-everything else
> > > remaining the same. At the end of the decade, the company's GR would
> > > be 613 million, or its valuation/capitalization would be almost 40%
> > > less. About half, for a consistent 10% inefficiency in management.
> > >
> > > Suppose the management of the company putzed around and didn't execute
> > > on engaging the marketplace for a year, then ran optimally. At the end
> > > of the decade, the company's GR would be 512 billion bucks, or its
> > > valuation/capitalization would be about half, for a 10% delay in
> > > execution.
> > >
> > > So, roughly, every delay or inefficiency by management of 10%
> > > ultimately costs the company 50% of its valuation/capitalization.
> > >
> > > John
> > >
> > > BTW, its not a linear relationship. Another inefficiency/delay of 10%
> > > doesn't cost another half billion bucks again. It cuts it in about
> > > half again.
> > >
> > > Note that a delay of one year costs a 50% market share in a
> > > competitive environment, so the market would shake out, ultimately, at
> > > a 2:1 market share ratio between a well run company, and a not so well
> > > run one, who has a 30% market share. In other words, a loss of 20%
> > > market share would spell the ultimate demise as the cost for the
> > > management mistake. It wouldn't make it through the decade.
> > >
> > > John Conover writes:
> > > > While we are on the subject of the dot-coms, (which is an interesting
> > > > case study-it is/was capitalism and free-market'ism at its finest-no
> > > > barrier to entry, insignificant infrastructure requirements, ample
> > > > investment or access to capital, and a choice of thousands for
> > > > consumers-what more could anyone want?) and completely mismanaged by
> > > > our finest young executives with their recent MBAs in tote. (Although
> > > > entropic methodologies are taught as a core competency requirement in
> > > > all university financial curriculum, it is not taught in business
> > > > school.)
> > > >
> > > > As an example, what is the chance that a company with 1% market share
> > > > will ever replace an industry leader with 50%?
> > > >
> > > > The answer is 2%, (you just divide the two numbers-its the gambler's
> > > > ultimate ruin scenario-and if it is an entropic system, it has to be
> > > > that way.)
> > > >
> > > > What that means is that, as an investor, one has to expect returns of
> > > > at least 50X to justify investing in companies with a 1% market share,
> > > > no matter how good the concept or vision is.
> > > >
> > > > We would, also, expect about a 1 in 10 survival rate, (e.g., any
> > > > company that has less than 10% market share, e.g., a 90% chance of
> > > > failure, won't make it,) and we are seeing those numbers.
> > > >
> > > > What's the duration of time one would expect for such a company to
> > > > be in business?
> > > >
> > > > About 50 * (50 - 10) = 2,000 days, or about 7.9 years for 253 business
> > > > days per year. (And, we have seen that, too; many of the dot-coms that
> > > > are failing were started as early as 1994, or so.)
> > > >
> > > > That's why so many dot-coms are failing.
> > > >
> > > > It is called management mistakes.
> > > >
> > > > John
> > > >
> > > > BTW, compare this to a company with competent executive staff that put
> > > > the company in biz early in the industry's development, started with,
> > > > and maintained a 50% market share. Such a company would be expected to
> > > > last about 100 * (100 - 50) = 5,000 days = 19.7 years, (at 253
> > > > business days per year.) The empirical metrics on the history of US
> > > > commerce in the 20'th century says 22 about years, (i.e., about a 10%
> > > > error in the probability estimate.)
> > > >
> > > > Only GE survived the entire century, BTW.
> > > >
> > > > But nothing is eternal in entropic systems. What's the chance of a
> > > > company in a leadership position eventually failing?
> > > >
> > > > Its a virtual certainty. That's the way capitalism works.
> > > >
> > > > John Conover writes:
> > > > >
> > > > > As approximate numbers, (taken from the average of many industries on
> > > > > http://www.johncon.com/ndustrix/,) to exploit the lack of
> > > > > predictability in the US GDP and its constituent industrial markets,
> > > > > the short term predictability of a few months to a 70% accuracy means
> > > > > that about (2 * 0.7) - 1 = 40% of a company's assets should be at risk
> > > > > in the short term, in WIP, capital expenditures, R&D, etc.
> > > > >
> > > > > Note that it is a conceptual framework, and just says what has to be
> > > > > done-not how to do it. The implementational paradigm is evaluation of
> > > > > "what if we do?", and "what if we don't?" scenarios, with more
> > > > > attention paid to the latter, (since you get more for better
> > > > > mitigation of risk than picking winners.)
> > > > >
> > > > > In the long term, (i.e., more than a few months,) mitigate risk
> > > > > through product diversification, (no less than 8 product lines, 10
> > > > > being about right, 12 probably too many,) and shuffle investment
> > > > > capital around in the product lines to avoid leptokurtotic behavior in
> > > > > the corporate P&L, (i.e., as a first order approximation, the GR
> > > > > generated by each product line should be about equal as a management
> > > > > paradigm; as a better approximation, the GR generated by each product
> > > > > should be proportional to the average divided by the deviation of the
> > > > > marginal increments in the GR. Likewise for the investment of
> > > > > capital.) That, also, defines investment in new product areas. The
> > > > > long term plan is to put about 2% of the company at risk, per business
> > > > > day, (or, cumulative, about 40% per month,) which is consistent with
> > > > > the short term strategy.
> > > > >
> > > > > The numbers would support a corporate growth of slightly more than 2X
> > > > > a year, and would maximize growth, while at the same time, minimizing
> > > > > risk exposure.
> > > > >
> > > > > Whether such growth could be managed and executed in the long run is
> > > > > an entirely different issue. (Its tough.)
> > > > >
> > > > > John
> > > > >
> > > > > BTW, there are other solutions for the same numbers-but they aren't
> > > > > optimally maximal. It is possible to grow a company faster than 2X a
> > > > > year-for awhile. Its a "coffin corner" solution, that will exhibit
> > > > > high growth, followed by a crash, (no matter what growth metric is
> > > > > used.) About 2X is the maximum SUSTAINABLE growth with those numbers.
> > > > > (Does the dot-com industry ring a bell as a case in point? Many of
> > > > > those companies exhibited growth of 2X in a single month-for
> > > > > awhile. The amount of the company placed at risk was far larger than
> > > > > the optimal 40% per month-most of it in capital investment. Glory has
> > > > > its price.)
> > > > >
> > > > > John Conover writes:
> > > > > > Interestingly, the lack of predictability in the US GDP is
> > > > > > exploitable.
> > > > > >
> > > > > > As a mathematical expediency most corporate strategies are divided
> > > > > > into short term and long term, (short term being a few months, long
> > > > > > term being more.)
> > > > > >
> > > > > > For long term strategies, the size of the window of predictability is
> > > > > > regarded as zero-meaning that the US GDP is treated as an entropic
> > > > > > system that behaves randomly. Strategies are developed that fit into a
> > > > > > framework determined by the average, (the potential gain,) and the
> > > > > > standard deviation, (the risk,) of the marginal increments of the US
> > > > > > GDP or industrial market-usually using monthly or quarterly data.
> > > > > >
> > > > > > Short term predictability is an inefficiency, (at least in the sense
> > > > > > of the efficient market hypothesis,) and can be exploited by adjusting
> > > > > > operations to near term GDP/market anticipations faster than anyone
> > > > > > else, by using the like of JIT techniques to minimize WIP risk, etc.
> > > > > >
> > > > > > The potential advantage is quite significant.
> > > > > >
> > > > > > John
> > > > > >
> > > > > > BTW, what would happen if every company that contributes to the US GDP
> > > > > > did that? Not much, except that it would grow faster. The US GDP would
> > > > > > become entirely entropic, (i.e., unpredictable,) and would be
> > > > > > efficient, and fair. It would be stable, (in the sense that it could
> > > > > > exist like that, forever.) Further, if the average of the marginal
> > > > > > increments equaled the variance, it would be maximally optimal.
> > > > > >
> > > > > > Unfortunately, it is a politically inexpedient solution. Such concepts
> > > > > > as monetary/fiscal policy to affect consistency in full employment,
> > > > > > etc., (which has been the paradigm of the last seven decades in all
> > > > > > industrialized countries,) would have to be discarded. Perceived lack
> > > > > > of influence over economic issues is not an expedient political
> > > > > > platform.
> > > > > >
> > > > > > (BTW, full employment is not necessarily optimal. It has, in general,
> > > > > > an unsustainable cost. Maximal sustainable employment is when the
> > > > > > GDP's average of its marginal increments equals the variance. However,
> > > > > > maximal employment does not, necessarily, mean full employment.)
> > > > > >
> > > > > > John Conover writes:
> > > > > > >
> > > > > > > In case you are curious, the big US economic recessions since
> > > > > > > Independence happened in 1819, 1833, 1837, 1857, 1873, 1893,
> > > > > > > 1929, (using the GNP/GDP numbers, which are not necessarily
> > > > > > > coincident with the stock market numbers, as far as downturns
> > > > > > > go,) and, (possibly-we don't know yet,) 2000.
> > > > > > >
> > > > > > > Note that this is the first generation in US history that has not
> > > > > > > had to endure a famine/depression, (at least yet,) and our
> > > > > > > perspective does not include how ugly they really are.
> > > > > > >
> > > > > > > John
> > > > > > >
> > > > > > > BTW, the numbers are interesting. To make predictions-like in the
> > > > > > > attached-the US GDP must be a deterministic system. Finding a
> > > > > > > mechanism that gives zero-free paths representing those numbers is
> > > > > > > a formidable proposition. If that can't be done, then, predictions
> > > > > > > can not be made.
> > > > > > >
> > > > > > > In NLDS systems, (of which the US GDP is certainly one,) the
> > > > > > > influence of the past on the future deteriorates rapidly-meaning that
> > > > > > > a small window into the past can be used to predict a small window
> > > > > > > into the future, and that is the best that can be done. The size of
> > > > > > > the window for the US GDP, is, at best, a few months, to a 70%, or so,
> > > > > > > accuracy.
> > > > > > >
> > > > > > > Unfortunately, the prevailing wisdom is that fiscal/monetary policy
> > > > > > > can not be used to influence the fluctuations in the US GDP-which
> > > > > > > was the paradigm of the past seven decades, and has since been
> > > > > > > abandoned.
> > > > > > >
> > > > > > > http://www.thestandard.com/article/0,1902,24243,00.html
--
John Conover, john@email.johncon.com, http://www.johncon.com/